fix(supervisor): bound the pod-count list server-side and clear its timer

Author: nicktrn

.server-changes/pod-count-exact-list.md

@@ -0,0 +1,6 @@
+---
+area: supervisor
+type: improvement
+---
+
+Self-hosted Kubernetes deployments now measure the running-task count exactly when deciding whether to pause pulling new work, instead of reading an approximate figure that could differ between reads. The safeguard now engages and releases at the point it is configured to, rather than slightly early or late.

apps/supervisor/src/clients/kubernetes.ts

@@ -69,9 +69,15 @@ export function createPodCountFetcher(
   namespace: string,
   timeoutMs: number
 ): () => Promise<number> {
+  const serverTimeoutSeconds = Math.max(1, Math.floor(timeoutMs / 1000));
+
   return async () => {
     const list = await withTimeout(
-      api.core.listNamespacedPod({ namespace, limit: 1 }),
+      api.core.listNamespacedPod({
+        namespace,
+        limit: 1,
+        timeoutSeconds: serverTimeoutSeconds,
+      }),
       timeoutMs,
       "pod count list"
     );
@@ -89,15 +95,21 @@ export function createPodCountFetcher(
   };
 }
 
-/** Rejects if `promise` outlives `timeoutMs`, so a hung request cannot freeze the caller. */
+/**
+ * withTimeout rejects if `promise` outlives `timeoutMs`, so a hung request cannot
+ * freeze the caller. A backstop only: the k8s client threads no AbortSignal through
+ * to fetch, so callers must also bound the request server-side (`timeoutSeconds`),
+ * or an abandoned request would stay open. The timer is cleared either way.
+ */
 function withTimeout<T>(promise: Promise<T>, timeoutMs: number, what: string): Promise<T> {
-  return Promise.race([
-    promise,
-    new Promise<never>((_resolve, reject) => {
-      setTimeout(
-        () => reject(new Error(`${what} timed out after ${timeoutMs}ms`)),
-        timeoutMs
-      ).unref();
-    }),
-  ]);
+  let timer: NodeJS.Timeout;
+  const deadline = new Promise<never>((_resolve, reject) => {
+    timer = setTimeout(
+      () => reject(new Error(`${what} timed out after ${timeoutMs}ms`)),
+      timeoutMs
+    );
+    timer.unref();
+  });
+
+  return Promise.race([promise, deadline]).finally(() => clearTimeout(timer));
 }